当用cmake编译的时候,出现了下面的错误:-- Check for working CUDA compiler: /usr/bin/nvcc -- broken ,我的电脑cuda和cudnn都是安装好了的,出现上面错误的原因可能是编译器找不到cuda或者sm_20和nvcc不兼容,解决办法,在CMakeLists.txt文件最上面添加下面代码: if (NOT CMAKE_CUDA_COMPILER) set(CMAKE_C...
2 changes: 1 addition & 1 deletion 2 ggml-cuda/fattn-vec-f32.cuh Original file line numberDiff line numberDiff line change @@ -149,7 +149,7 @@ static __global__ void flash_attn_vec_ext_f32( for (int i0 = 0; i0 < D/2; i0 += WARP_SIZE) { const int i = i0 + ...
This was broken in 2.0 but is fixed in the nightly (torch-2.1.0.dev20230402+cu118) for me. Activation checkpointing uses more VRAM than without, but that's probably a separate issue. nyngwang reacted with thumbs up emoji 👍 Sorry, something went wrong. ...
On the fly syntax checking for GNU Emacs. Contribute to flycheck/flycheck development by creating an account on GitHub.